Causal Associations Between Plasma Protein Levels and Alzheimer Disease

Abstract

AbstractBackgroundThe aim of this study was to infer causal relationships between protein levels measured in blood and the risk of late‐onset Alzheimer disease (AD).MethodTwo‐sample Mendelian randomization (MR) was used to infer the causal effect of protein levels on AD risk. Summary statistics from genome‐wide association studies (GWAS) for 4,907 SomaScan aptamers (measuring 4,719 proteins) from Ferkingstad et al. (2021) in 35,559 Icelanders (average age = 55±17years), and for AD GWAS from Bellenguez et al. (2022). Instruments for exposures were selected at p < 5×10−8 and p < 5×10−6 thresholds respectively for the main and sensitivity analyses. Instruments were pruned for linkage disequilibrium (r2 < 0.0001) to ensure selection of independent instruments. Because APOE is highly pleiotropic and exposure effect size is hypothesized by MR not to be larger than the outcome (Steiger filtering), thus variants within ±1MB from APOE were removed in the MR‐analysis. Two tests were considered: Inverse‐Variance‐Weighted and Weighted‐Median, which allows robust MR estimates even when some invalid instruments are included.ResultAmong known AD‐GWAS genes, TREM2, PILRA, ACE, CR1, and BIN1 protein levels were associated with AD. Proteins without prior AD‐GWAS links were also found to be associated with AD including: PPBP, NRP2, AZGP2, and TNFSF13B/BAFF (Figures 1&2). Interestingly, APOE and KLOTHO levels were not associated with AD (Figures 3&4), though increased APOE level significantly associated with increase AD risk when including cis‐APOE variants at p < 5×10−6 (Figure 5).ConclusionOur study informs the directionality of association between certain plasma protein levels and AD risk, and prioritizes candidate proteins to be considered for AD drug development.